Automated ad creation with Claude Fable 5.
Automated ad creation turns a brief into copy, creative, and campaign structure — then refreshes it as performance decays. Here are the eight workflows and copy-paste prompts that run it, and the line where the automation should stop.
Built by our community of 2,000 marketers
Free skills and prompts for paid ads and SEO
Templates for Claude, ChatGPT and Perplexity.
Clients we work withClients we
work with
Before you read · 30-second setup
Connect Claude to your ad accounts in one click
Every prompt and skill below works best against live data. Copy the Ryze MCP link, open Claude → Settings → Connectors → Add custom connector, paste, and authenticate Google Ads or Meta with OAuth. Done — Claude can now read your real campaigns.
https://connector.get-ryze.ai/mcp
Inside Ryze AI: the same MCP link on the Home screen — copy it there or right here, it’s the same connector.
What Claude builds once it’s connected — demo snapshots
Google Ads account audit — health score, severity findings, scorecard.
30-day performance audit — ROAS, channel split, lost impression share.
Meta creative fatigue — frequency, CTR decay, refresh-now matrix.
Paid funnel diagnostic — GA4 + ads, drop-offs, recoverable revenue per page.
Automated ad creation is the use of AI to generate ad copy, image and video creative, and campaign structure from a brief — and then to regenerate them on a schedule as performance decays.
In 2026 the bottleneck has moved. Producing forty ad variants is no longer the hard part; deciding what forty variants should say still is. The workflows below split the job accordingly: the model does the volume work — briefs, variants, resizing, naming, fatigue detection — and a human keeps the offer, the angle, and the approval before spend. Treat it as a production line you supervise, not a button that makes ads.
The four-stage automated ad creation workflow
Brief — mine the angles that already work, then write the brief
Generate — copy, image and video concepts against a real offer
Assemble — campaign structure, budgets, naming, and UTMs
Refresh — catch creative fatigue and regenerate what's dying
Each stage below is a prompt you can paste into Claude and schedule. Connect the Ryze MCP once so Fable 5 reads your live Google Ads and Meta data, and the diagnostic stages run against real numbers instead of guesses.
- Stage 1: Brief — what to say, sourced from what already works
- Stage 2: Generate — copy and creative concepts, on brand
- Stage 3: Assemble — structure, naming, and tracking that survives reporting
- Stage 4: Refresh — fatigue detection and regeneration
What automates, and what still needs a human
The useful question isn’t whether AI can make an ad — it can. It’s which parts of the job get faster without getting worse. Typical production time for one campaign’s creative set, manual versus automated:
| Step | Manual | Automated | Human still owns |
|---|---|---|---|
| Angle research | 2–4 hours | 10 minutes | Which angle to bet on |
| Ad brief | 1–2 hours | 10 minutes | The offer itself |
| Copy variants (×20) | 2–3 hours | 5 minutes | Brand voice sign-off |
| Static creative set | 4–8 hours | 20–40 minutes | Art direction |
| Video concepts | 1–3 days | 2–4 hours | Final edit and taste |
| Campaign build | 1–2 hours | 15 minutes | Budget and bid strategy |
| Fatigue detection | Weekly, by hand | Scheduled, daily | The decision to kill |
Read the right-hand column as the point of the exercise. Automation collapses production time; it doesn’t transfer judgment. The accounts that get worse after adopting AI creative are usually the ones that automated the approval step too. For the copy half specifically, see ChatGPT prompts for ads.
The prerequisite: input quality
Automated ad creation fails in a predictable way: fed a thin brief, it produces forty variants of the same generic sentence, and the account gets slower instead of faster. Every prompt below assumes four inputs are real. Fix these first and the rest of the workflow works.
A specific offer
Not "quality products" — a named product, a price, and a reason to act now. The model can phrase an offer; it cannot invent one that's true.
A named audience
Who it's for, what they currently do instead, and the objection that stops them. Vague targeting produces vague copy.
Current brand voice samples
Three to five ads or pages that sound right. Without samples the output defaults to generic marketing register.
Live performance data
What already works in the account. Angle mining against your own winners beats angle mining against the open web.
00 · Brand Input Auditor
Fable 5Is my brief good enough to automate from?
1 · Brief — angles and the ad brief
Generation quality is decided before a single headline is written. These two workflows produce the brief the rest of the chain runs on — one mines the angles that already convert, the other turns a product page into a structured brief a model can execute against.
01 · Winning-Angle Miner
Fable 5Which angles already work here?
Reads your own top performers and the messaging patterns in your category, then ranks the angles worth testing next — so generation starts from evidence, not from a blank page.
02 · Offer & Audience Brief
Turn a product page into an ad brief
Converts a URL into the structured brief every downstream prompt expects: offer, audience, objection, proof, and the one sentence the ad has to land.
2 · Generate — copy and creative
With a real brief in hand, generation is the cheap part. The two prompts below produce platform-shaped copy and creative concepts that a designer or an image model can actually execute — concrete enough to shoot, not mood-board abstractions.
03 · Ad Copy Generator
Fable 5Headlines and primary text, on brand
Generates variants that respect character limits and your actual voice, each tagged with the angle it tests — so the results tell you something when they come back.
04 · Creative Concept Brief
Image and video concepts, spec'd to shoot
Turns each winning angle into a concrete visual brief — subject, composition, on-screen text, and first frame — specific enough to hand to a designer or an image model.
One rule worth enforcing in every generation prompt: no claim the brief doesn’t support. Left unconstrained, models fill gaps with plausible-sounding specifics — invented percentages, awards you didn’t win, guarantees you don’t offer. That’s a compliance problem on Meta and Google, not a style one. Constrain the claims, then review before spend.
Get a free instant audit
Get a free, instant read on your paid ads or SEO — and fix it right away.
Paid ads audit
- Catch wasted spend & broad-match leaks
- Find account structure gaps
- Rank your quickest wins
- Spot PMax & brand-search overlap
- Check conversion-tracking health
- Benchmark CPC vs your industry
- Catch wasted spend & broad-match leaks
- Find account structure gaps
- Rank your quickest wins
- Spot PMax & brand-search overlap
- Check conversion-tracking health
- Benchmark CPC vs your industry
Free · no credit card · instant
SEO audit
- Find keyword & ranking gaps
- Catch technical SEO issues
- Rank your fastest wins
- Surface thin & duplicate pages
- Check indexing & crawl coverage
- Compare backlinks vs competitors
- Find keyword & ranking gaps
- Catch technical SEO issues
- Rank your fastest wins
- Surface thin & duplicate pages
- Check indexing & crawl coverage
- Compare backlinks vs competitors
Free · no credit card · instant
3 · Assemble — structure, naming, tracking
Generated assets are worth nothing until they’re shipped in a structure you can read later. This is the stage teams skip, and it’s why so many AI-generated campaigns produce volume no one can attribute. Two prompts: one builds the structure, one enforces the naming.
05 · Campaign Assembler
From approved brief to campaign structure
Maps approved variants onto ad sets and audiences with a testing structure that can actually reach significance — rather than 30 ads splitting a budget forty ways.
06 · Naming + UTM Builder
Fable 5Naming that survives reporting
Generates a consistent naming convention and matching UTMs across every asset, so angle-level performance is queryable three months from now instead of guesswork.
If you’re pulling creative performance programmatically, the field-level detail is in the Meta Marketing API creative performance guide.
4 · Refresh — creative fatigue
Automated creation earns its keep here. Producing ads faster only pays if you also replace them faster, and creative fatigue is a scheduling problem — frequency climbs, CTR decays, CPM drifts up, and the account quietly gets more expensive. This one runs daily and tells you what to regenerate.
07 · Creative Fatigue Watchdog
What’s dying, and what should replace it
Checks every active ad for the fatigue signature, ranks what to retire, and drafts replacements built on the angle that’s still working — so the refresh queue never runs dry.
Related reading: Claude Fable 5 for Google and Meta Ads covers the bidding and budget half of the same account, and the social media marketing costs guide covers what production actually costs when it isn’t automated.
Setup: connect your ad accounts
The Angle Miner, Campaign Assembler, and Fatigue Watchdog all need live data. One MCP URL gives Fable 5 Google Ads and Meta access; paste it into Claude once and every scheduled prompt runs against current numbers.
Install Claude Desktop
Download from claude.ai. Free plan works; Pro unlocks longer scheduled runs. Fable 5 is the default model.
Paste the MCP URL
Settings → Connections → Add MCP Server. Paste the Ryze Ads Command Center URL. Connect.
Authenticate
OAuth for Google Ads and Meta — one click each, under 30 seconds. Every prompt above now has live data.
Prefer it run itself? Ryze AI is the autonomous platform that executes this loop 24/7 rather than only recommending it — mining angles, generating and rotating creative, assembling campaigns, and retiring fatigued ads across Google, Meta, and ChatGPT without manual prompting. Free tools for the individual steps live in the Ryze tools suite.
Frequently asked questions
What is automated ad creation?
Automated ad creation is the use of AI to generate advertising assets — copy, image and video creative, and campaign structure — from a brief, and to regenerate them on a schedule as performance decays. It covers the production line from angle research through creative refresh. It does not cover the offer, the strategic angle, or the approval before spend, which remain human decisions.
Can AI create ads without any human input?
Not well. The models are reliable at volume work — variants, resizing, naming, briefs, fatigue detection — and unreliable at deciding what is worth saying. Left unsupervised they also invent claims: percentages, awards, guarantees. Both Meta and Google treat unsubstantiated claims as a policy issue, so an approval step before spend is a practical requirement, not caution.
What does automated ad creation actually speed up?
Mostly production time. A creative set that takes 8 to 14 hours to produce manually — angle research, brief, copy variants, statics, campaign build — typically lands in under two hours automated. Video is the exception: concepts generate quickly, but a finished edit still needs human judgment on pacing and taste.
Which parts should stay manual in 2026?
Four: the offer itself, the choice of which angle to bet budget on, art direction, and the decision to kill a campaign. Each is a judgment call with real money attached, and each is where models are weakest. Automating the approval step is the most common way accounts get worse after adopting AI creative.
How do I stop AI-generated ads from sounding generic?
Feed it real inputs. Three to five brand voice samples, a named audience with a stated objection, and a concrete offer with a price. Generic output is almost always a thin-brief problem rather than a model problem — the Brand Input Auditor prompt above exists to catch it before you generate.
How often should automated creative be refreshed?
Watch the signature rather than the calendar: rising frequency with falling CTR, or CPM climbing more than about 20% at a flat conversion rate. Run the fatigue check daily and refresh when an ad trips it. Fixed rotation schedules retire winners early and keep losers alive.
Does automated ad creation work on Google and Meta both?
Yes, though the shape differs. Meta rewards creative volume and angle diversity, so generation pays off fastest there. Google Ads — especially Performance Max — absorbs assets into its own combinations, so the value is in feeding it strong, varied inputs and disciplined naming rather than in producing more finished ads.
Where do I schedule these prompts?
In Claude Cowork → Scheduled: paste the prompt, name it, pick a frequency, save. Each card above lists its suggested cadence — angle mining monthly, generation per campaign, the fatigue watchdog daily. Keep every scheduled run read-only until you trust its output.
